New functions to perform a precheck for the user data and metadata
parameters that are going to be used to execute a OBS_GetData operation.
Doing this before the OBD_GetData we are able to check if we're going to
have problems with the current request. Is useful because right now
we're using the *_Exception_Safe functions which return NULL in case we
have an exception, this leads to execute the whole request even if we
know beforehand that it's going to fail.
I was creating the wrappers before the wrapped function and postgres
extensions do not support that.
I did not notice in my local environment cause the old functions were
already there.
As shown in the tests, that type has the flexibility of being able to be
cast to from any other numeric type. This gives us flexibility without
bloating the API with lots of type-specific functions.
- Fixed fallback becuase we have to use spiexceptions in this case
because we retrieve the configuration usig plpy.execute fuctions that
wrap any exception into a spiexception.
- In case we don't want to use Mapzen, we could leave the api_key empty
becuase we arise a ConfigException if the Mapzen api_key is empty so
we are able to make fallback. Right now we can't remove the mapzen
configuration because it'll fail when the InternalGeocoderConfig try to
load the ServiceDBConfig
This is a dirty hack, we should improve how the DB config is loaded. See
Two problems fixe with this functions ported from the cartodb extension:
* There was an incorrect reference to the cartodb scchema
* They need to be SECURITY DEFINER to be usable with the geocoder_api role
ServiceManager class has been introduced to handle service configuration at
SQL level (with a LegacyServiceManager alternative for non-refactored services).
These new classes take the responsibility of rate limits and quota checking.
Tests have been added for ServiceManager and rate limits, but currently they
check only the limits configuration since Lua support would be needed
to use rratelimit with MockRedis.
Instead of getting (deleted and untracked files):
```
$ git status
# On branch master
# Your branch is ahead of 'origin/master' by 20 commits.
#
# Changes not staged for commit:
# (use "git add/rm <file>..." to update what will be committed)
# (use "git checkout -- <file>..." to discard changes in working directory)
#
# deleted: cdb_dataservices_server--0.20.0--0.21.0.sql
# deleted: cdb_dataservices_server--0.21.0--0.20.0.sql
# deleted: cdb_dataservices_server--0.21.0.sql
# modified: cdb_dataservices_server.control
#
# Untracked files:
# (use "git add <file>..." to include in what will be committed)
#
# cdb_dataservices_server--0.21.0--0.22.0.sql
# cdb_dataservices_server--0.22.0--0.21.0.sql
# cdb_dataservices_server--0.22.0.sql
# old_versions/cdb_dataservices_server--0.20.0--0.21.0.sql
# old_versions/cdb_dataservices_server--0.21.0--0.20.0.sql
# old_versions/cdb_dataservices_server--0.21.0.sql
no changes added to commit (use "git add" and/or "git commit -a")
```
you'd get something like:
```
$ git status
# On branch master
# Your branch is ahead of 'origin/master' by 20 commits.
#
# Changes to be committed:
# (use "git reset HEAD <file>..." to unstage)
#
# renamed: cdb_dataservices_server--0.20.0--0.21.0.sql -> old_versions/cdb_dataservices_server--0.20.0--0.21.0.sql
# renamed: cdb_dataservices_server--0.21.0--0.20.0.sql -> old_versions/cdb_dataservices_server--0.21.0--0.20.0.sql
# renamed: cdb_dataservices_server--0.21.0.sql -> old_versions/cdb_dataservices_server--0.21.0.sql
#
# Changes not staged for commit:
# (use "git add <file>..." to update what will be committed)
# (use "git checkout -- <file>..." to discard changes in working directory)
#
# modified: cdb_dataservices_server.control
#
# Untracked files:
# (use "git add <file>..." to include in what will be committed)
#
# cdb_dataservices_server--0.21.0--0.22.0.sql
# cdb_dataservices_server--0.22.0--0.21.0.sql
# cdb_dataservices_server--0.22.0.sql
```
which is nicer IMHO.
- It's not mandatory to have the log_path defined for all the services
- Now we add metrics to a MetricsGatherer using the add method in order
to gather data from all the function scope, not just in one defined
point
- There is a `metrics` context manager to wrap the function block. This
context manager get the execution time for the block among other
things
